RT CUDA Activeshade errors with 3.6

[simona.b]

Hi,

I've been having a lot of different errors whilst in Activeshade since upgrading to Vray 3.6 (using RT CUDA), not sure if this is linked to the new version or perhaps because of a complicated 3ds Max scene (has xref scene objects). Maybe I'm running out of graphics card memory, hard to tell as I would be rendering using Activeshade, moving around the object ok, then suddently it would error out and stop. The stop button inside vray frame buffer then often becomes unresponsive if I try to render again and doesn't allow me to shut 3ds Max down, but I can still save and manouver around the UI, just have to force end task to close software.

Happened so many times now on this particular file, but I cannot share the scene file due to it being a military project for now.
I have tried hybrid rendering by default, but also with just graphics cards and same problem keeps happening.


My PC spec:
2 x nvidia gtx 1080
drivers: 378.92

windows 10 , 64bit
i7-6900k cpu
RAM 32gp
3ds max 2017 : 19 SF


some of the errors I have encountered:
vray error unhandled exception c0000005
cuda error 719

I have attached some of the errorlogs and dmp files from today.

Hope it's possible to shed some light why this is happening. (probabaly user error!)

kind regards,
Simona

[savage309]

It seems like bug at our side (from the dump file it seems related with the anisotropy).
It would be really helpful if we have the scene file.

Best,
Blago.

[simona.b]

Thanks Blago for taking a look at the dump files for me, really appreciate it.

Just tested the same scene on a colleagues's pc, using vray 3.5 - cannot replicate the crash. So might have to revert back a version of vray on my pc.
Tried replicating the scene using the object with anisotropic material, no crashing. Tried deleting almost everything from the scene and only leaving a few objects, again, no crash. Wanted to create a scene that would crash for your debugging purposes, but cannot replicate it only in this large scene full of xref scenes, some xrefs inside xref.... So maybe Activeshade GPU is working a bit hard and ocasionally desides to overdo it? it seems to happen when I drag the animation timeline slider up, sometimes it updates fine, othertimes it errors out, as if it ran out of GPU memory because the cache or something was full? This is all me guessing though!

I guess for now I just won't use activeshade too much, as everything seems to render fine if I'm careful, just fustrating if it does crash, it forces me to shut down max as it no longer detects graphics cards for restarting the render.

Unfortunatelly I wasn't able to get permission to upload the scene file, however if I do encounter something similar again in testing "teapot" scenes, I will do.

Kind regards,
Simona

[savage309]

Thanks, we will try to replicate that here.
Sorry for the trouble.

Best,
Blago.

[JD3D_CGI]

I have the same problem. Would it help to have another scene that is giving you this problem?
Strangely enough - it worked fine last week, and I only made basic changes - changes that had rendered fine before and now I keep getting this error!

[savage309]

Yes, having the scene will help (at least we can confirm that it is the same anisotropy related problem).

Best,
Blago.